kompute : fix fallback to CPU (#5201)

diff --git a/llama.cpp b/llama.cpp
index 9631506c6fd4e9499f7b4efa8aecbeb680454996..796aaa895c05b6b7433efa88900a4379d7b06f13 100644 (file)
--- a/llama.cpp
+++ b/llama.cpp
@@ -4136,7 +4136,7 @@ static int llama_model_load(const std::string & fname, llama_model & model, llam
         }
 
 #ifdef GGML_USE_KOMPUTE
-        if (ggml_vk_has_device() && params.n_gpu_layers > 0 && (
+        if (params.n_gpu_layers > 0 && (
             !(model.arch == LLM_ARCH_LLAMA || model.arch == LLM_ARCH_FALCON)
             || !(
                 model.ftype == LLAMA_FTYPE_ALL_F32 ||
@@ -4145,8 +4145,8 @@ static int llama_model_load(const std::string & fname, llama_model & model, llam
                 model.ftype == LLAMA_FTYPE_MOSTLY_Q4_1
             )
         )) {
-            // disable Vulkan due to unsupported model architecture or quantization type
             // TODO(cebtenzzre): propagate this error outside of llama_load_model_from_file
+            LLAMA_LOG_WARN("%s: disabling Kompute due to unsupported model arch or quantization\n", __func__);
             params.n_gpu_layers = 0;
         }
 #endif